Aerosol Optical Depth over Land. The aerosol optical thickness products over land are accurate to within their calculated uncertainties (±0.05±0.20*tau), except in situations with possible cloud contamination, over surfaces with subpixel surface water such as coastal areas and marshes, and over surfaces with subpixel snow or ice cover.

Aerosol Optical Depth over Ocean. The aerosol over ocean optical thickness products are accurate to within their calculated uncertainties (±0.03±0.05*tau) at wavelengths at 660 nm and 870 nm. In dust aerosol regimes, retrieved aerosol optical thickness at other wavelengths will have greater error due to nonspherical effects. Larger errors may also occur in situations with possible cloud contamination. This cloud contamination appears to occur more frequently in high latitudes above low altitude ice clouds.

Retrievals over Glint Regions. All aerosol products retrieved from within the sun glint mask of 40 degrees are to be considered experimental in nature and thus have quality flag set to 0 (no confidence).

Information provided by Dr. Yoram Kaufman & Dr. Lorraine A. Remer, Code 913, NASA/Goddard Space Flight Center.


Water Vapor (05_L2)

Clear-sky Precipitable Water (Near Infrared). No problems noted.   It is interesting that MODIS near-IR precipitable water results compared well with DOE ARM microwave radiometer water vapor over an extended time period (about 1 year). There was also excellent agreement with microwave radiometer data collected in a recent China field experiment.


Cloud (06_L2)

Cloud Optical Thickness & Effective Radius. Retrievals with anomalously small optical thicknesses and large effective radii can result from partially cloudy pixels. This is especially noticeable for broken cloud fields over dark surfaces, for example, broken marine stratocumulus and trade cumulus scenes.

Information provided by Dr. Steven Platnick, Code 913, NASA/Goddard Space Flight Center.

Cloud Top Pressure & Temperature. The MODIS cloud-top pressure/temperature/height product relies on IR sounding bands located near the 15 µm CO₂ band. The algorithm is fairly robust and operates on both daytime and nighttime data. The user should be aware of the following problematic areas. Areas of difficulty include:

• Optically thin cirrus. When the cloud signal, i.e., the difference in radiance between the cloud and clear-sky, is low, and thus comparable to instrument noise, it is difficult to assess cloud pressure.
• Multilayered clouds, specifically when thin cirrus overlies a lower-level water cloud. The cloud height assessment is performed operationally under the assumption that only a single cloud layer exists in the field of view. Biases occur in cloud height when more than one cloud layer exists and the uppermost layer is optically thin.
• The inference of cloud height in a relatively isothermal atmosphere (e.g., polar regions). The CO₂ slicing algorithm requires a temperature gradient in its implementation.
• Low-level clouds. Because of signal-to-noise considerations, the CO₂ slicing algorithm is applied only to clouds at pressures lower than 700 hPa. Low-level cloud heights are based solely on the 11 µm band under the assumption that the cloud is optically thick.

Cloud Fraction (Night). Cloud Fractions for nighttime data are too high over the oceans. This is an artifact of the cloud mask. In the real world, there is a small increase in clouds over ocean at night, but not to the extent seen in the Cloud Fraction Night product. The main cause is the use of a strict 11 µm variability test, which is used at night, but not during the day. This is done in an effort to compensate for the absence of visible data in detecting low-level clouds. This has the effect of extending clouds outward 1 pixel from the actual cloud edge and the result is an increase in cloud fraction in areas characterized by scattered or broken cloud fields. Steps are being taken to remedy the situation as much as possible in an upcoming Collection 5 delivery by using sea surface temperature maps and observed clear-sky brightness temperature maps from cloud mask processing on the previous 8-days.

Information provided by Richard Frey, University of Wisconsin


Profile (07_L2)

Clear-sky Precipitable Water (Infrared). No problems noted.


Cloud Mask (35_L2)

Cloud Mask. Issues expected to be improved with next algorithm update:

• Some false clouds indicated over coastlines, rivers, and inland lakes
• 250 meter cloud mask should be used with caution
• False clouds indicated over some desert regions at night
• Aerosol detection (smoke, fire, dust) needs improvement

• Algorithm does not work well in regions of Polar night
• Shadow detection is problematic